Mining and interpretation of association rules among protein sequence motifs

In research of molecular biology, it is essential to characterize proteins functionally. One of such characterization method is to predict function from its amino acid sequence. Motifs are representative features of functions in a protein. However, little is known about relations in more than two motifs. In this paper, we will find association rules on motifs using data mining algorithm. Our result shows about 1,300,000 rules with lower bound of 95% confidence and 0.003% support. The rules have enriched biological meanings. As an example of pattern interpretation, we examine two intracellular Ca²⁺-release channels: ryanodine receptor and inositol 1,4,5-trisphosphate receptor.